On Thu, Jul 28, 2011 at 19:42, Jake Penton
h:: a h = 'a'
to which ghci replies:
Couldn't match type `a' with `Char' `a' is a rigid type variable bound by the type signature for c :: a at /Users/David/Project/EoP/ch04/weak.hs:114:1 In the expression: 'a' In an equation for `c': c = 'a'
This last example is probably the most basic one which I need to understand. But, why is the problem apparently a different one than in the definition of "f" above?
The other one was complicated by polymorphism: a numeric literal is compiled into a call to fromIntegral, whose result type is Num a => a. The problem is that, when you say something's type is "a", you are not saying "pick an appropriate type for me"; you are saying "whoever invokes this can ask for any type it wants" (equivalently: "I promise to be able to produce *any possible* type"). But then you give the value as Num a => a in the first example and Char in the second example, neither of which is "any possible type". An explicit type is a complete contract; having contracted to produce an "a" (any type), you can't then offer only a Char or a Num a => a. You have to satisfy the contract which says "any type", otherwise you're doing the type checking equivalent of a bait-and-switch. You can't express "pick a type for me" in a type signature; types are concrete, and a type variable in a signature is a concrete "anything", meaning the caller can request whatever it wants and you must produce it. The type must be *completely* described by the signature; what it says is what you're committed to, and you can't then offer something else. If you need a partial type signature, there are some tricks you can use which let you force types in the implementation without specifying a concrete signature (see http://okmij.org/ftp/Haskell/types.html#partial-sigs). -- brandon s allbery allbery.b@gmail.com wandering unix systems administrator (available) (412) 475-9364 vm/sms